When Mike Goldman offered $5,000 to anyone who could compress a randomly-generated file of their chosen size in early 2001, he thought his money was perfectly safe. After all, it's mathematically impossible to compress truly random data. But Patrick Craig spotted a loophole that would lead to one of the most entertaining — and ultimately unpaid — exchanges in compression challenge history.
The key was in Goldman's March 2001 casual agreement to allow "several compressed files" rather than just one. Craig requested a 3MB test file and split it into 219 pieces, removing a single character from where the pieces joined.
Craig submitted his work to Goldman, and wrote: "The total file size of the decompressor (156 bytes) and all the compressed files (3,145,528 bytes) is 3,145,684 bytes. The original file is 3,145,728 bytes, so I successfully completed the challenge."
His "decompressor" simply reassembled the pieces and reinserted the missing character. Technically, the total size of his files was 44 bytes smaller than the original — meeting the letter, if not the spirit, of Goldman's challenge.
"The Challenge had nothing to do with compressing data," Craig argued when Goldman refused to pay. "It was simply to recreate the original file from several 'compressed' files and a 'decompressor' whose total file size was less than the original file."
Goldman threatened legal action, arguing that filesystem tricks didn't constitute real compression: "EOF does not consume less space than '5', no compression has occurred, data has been shifted or 'hidden' from the file itself to another part of the filesystem."
The resulting debate on comp.compression sparked intense discussion about what truly constitutes compression. As one commenter noted, "Who was trying to trick who? Wasn't Mike trying to trick some naive person into accepting a challenge that couldn't be met? So Patrick out-tricked the tricker…. that's the thing that great tales are made of."
Craig ultimately withdrew his claim to the $5,000, letting Goldman keep his initial $100 entry fee. His real goal was to expose the danger of carelessly-worded challenges — and perhaps teach a lesson about the perils of trying to trick "naive" challengers.
Previously:
• The sad life and death of Phil Katz, creator of the Zip file compression program
